GPU runner

NVIDIA A100 40GB runners for GitHub Actions

The workhorse training and inference GPU. Available on AWS, GCP and Azure with mature CUDA + NCCL tooling. Cirun spins NVIDIA A100 40GB VMs up on your cloud account per workflow — pick whichever cloud has the SKU on your existing account, CUDA + drivers come from each cloud's own Deep Learning image.

Run a workflow on NVIDIA A100 40GB Docs

Why this fits

40 GB VRAM and ~312 TFLOPS FP16 — mid-size LLM evals and SDXL-class generation.
Works with PyTorch, JAX, TensorFlow — CUDA + drivers come from your cloud's official Deep Learning AMI / image, not your workflow.
Ephemeral by default — each job gets a clean VM, so leaked weights or hung CUDA contexts can't poison the next run.

.cirun.yml

1runners:
2  - name: gpu-a100-40gb
3    cloud: aws
4    instance_type: p4d.24xlarge
5    # Use AWS Deep Learning AMI GPU PyTorch on Ubuntu 22.04, or the
6    # Cirun-published NVIDIA AMI.
7    machine_image: ami-04823729c75214919
8    labels:
9      - cirun-gpu-a100-40gb

.cirun.yml

1runners:
2  - name: gpu-a100-40gb
3    cloud: gcp
4    instance_type: a2-highgpu-1g
5    # Use a GCP Deep Learning VM image family
6    # (deeplearning-platform-release) for pre-installed CUDA + drivers.
7    machine_image: projects/deeplearning-platform-release/global/images/family/pytorch-latest-cu121
8    labels:
9      - cirun-gpu-a100-40gb

.cirun.yml

1runners:
2  - name: gpu-a100-40gb
3    cloud: azure
4    instance_type: Standard_ND96asr_v4
5    # Use Microsoft's HPC/DSVM Ubuntu image for pre-installed CUDA +
6    # drivers.
7    machine_image: microsoft-dsvm:ubuntu-hpc:2204:latest
8    labels:
9      - cirun-gpu-a100-40gb